I don't think we need a separate local_install target, as we can just pass the destination directory on the make command.
It would still be convenient, and it would help enforce some kind of de-facto standard location for locally installed modules, so that one reasonably can assume that it's, say, $HOME/pike-modules and not $HOME/pike/local_modules or $HOME/PikeStuff or whatever. If nothing else, it's useful to have a common term to talk about.
Speaking of the destination, what's the most reliable way of finding the system module directory?
The same way Tools.Standalone.module finds out the include directory. I.e. you should extend the master to store lib_prefix in a separate variable. Preferably it should be possible to query it too with pike -x module.
I see there's a share_prefix too which is intended for architecture independent data, i.e. undumped modules. Can't find any installation method that installs there, though.
/ Martin Stjernholm, Roxen IS
Previous text:
2003-10-30 06:10: Subject: Re: module build process
The install target doesn't seem to be used during the build/install of pike proper. I will therefore make the bold statement that it should be used during the build and installation of external modules, and will proceed to rework it with that in mind. I don't think we need a separate local_install target, as we can just pass the destination directory on the make command.
Speaking of the destination, what's the most reliable way of finding the system module directory? master()->pike_module_path[-1] seems to be the right location when PIKE_MODULE_PATH is also set. Anyone have information to the contrary?
Bill
/ Brevbäraren